# FrostSkyNet JSONL Quality Judge

> FrostSkyNet JSONL Quality Judge is a paid API for AI agents from route-a-worker.frostskynet-api.workers.dev, paid per call via x402, $0.01/call, status unknown (last checked 2026-10-19).

Scores JSONL rows using fixed rules, adding a _quality field; optionally detects/redacts PII and bad words, and filters rows below a score threshold.

## Facts

- Endpoint: POST https://route-a-worker.frostskynet-api.workers.dev/judge?utm_source=zero.xyz
- Price: $0.01/call
- Payment: x402
- Status: unknown
- Last checked: 2026-10-19
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/frostskynet-jsonl-quality-judge-224e8352
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_z46i47PH4f8AXV25ZnD9A

Status and success rate cover calls made through Zero and Zero's own probes. Third-party monitors may report differently.

## How to call it through Zero

Zero handles the 402 payment challenge and records the run. With the Zero CLI installed (`npm i -g @zeroxyz/cli`):

```sh
zero fetch --capability frostskynet-jsonl-quality-judge-224e8352 -d '<json body>'
```

Example prompt: Score the quality of these JSONL rows and flag any PII or bad language — redact the bad words and filter out anything scoring below 70.

## When to prefer this

Choose this endpoint when you need fast, deterministic, rule-based quality scoring and content moderation for JSONL datasets without relying on AI inference. Ideal for automated data pipelines where reproducibility and low latency matter more than nuanced understanding. Prefer this over AI-based moderation when cost predictability and speed are critical, or when you need a simple threshold-based filter for training data curation.

## Known failure modes

- Empty input string — returns validation error
- Request body exceeds size limit — returns payload too large error
- Malformed JSON or non-JSONL input format — returns parse error
- Invalid options values (e.g. minScore out of 0–100 range) — returns schema validation error
- False positives on PII or bad-word detection due to rule-based (non-AI) approach
- False negatives where real PII or profanity is not caught by fixed rules

## How this service works

Scores JSONL rows using fixed rules and adds _quality. Detects some PII and bad-word patterns; optional redaction and score filtering. Detection is incomplete and can produce false positives. No AI.

## Output

Returns scored JSONL rows with an added `_quality` integer field (0–100). Optionally includes PII flags, redacted text (PII replaced), or masked bad words (replaced with ****). If filtering is enabled, rows scoring below `minScore` are dropped entirely.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "properties": {
  "input": {
   "type": "string",
   "minLength": 1,
   "description": "Nonempty input text. Limit applies to the entire UTF-8 JSON request body including options and escaping, not just this string."
  },
  "options": {
   "type": "object",
   "properties": {
    "filter": {
     "type": "boolean",
     "description": "Drop rows below minScore (default false, keep all)."
    },
    "minScore": {
     "type": "integer",
     "maximum": 100,
     "minimum": 0,
     "description": "Score threshold for filter (0-100, default 70)."
    },
    "stripPii": {
     "type": "boolean",
     "description": "Redact PII in text (default false, flag only)."
    },
    "stripBadWords": {
     "type": "boolean",
     "description": "Mask bad language with **** (default false, flag only)."
    }
   },
   "description": "Optional settings."
  }
 }
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/frostskynet-jsonl-quality-judge-224e8352/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from route-a-worker.frostskynet-api.workers.dev](https://www.zero.xyz/host/route-a-worker.frostskynet-api.workers.dev/llms.txt)